Introduction Materials and methods Experimental data

Experiments from 2 to 5 aimed all at evaluating bispecific one legume and one non-legume mixtures under actual conditions (water and N limiting conditions). For all these experiments, no fertilizer N was added and irrigation was provided only at sowing in Experiments 2, 4 and 5 to ensure optimal plant emergence and obtaining homogeneous plant stands. No irrigation was used afterwards during the cover crop growth season leading to the occurrence of water stress. In Experiment 3, no irrigation occurred at any time because sufficient rainfall ensured optimal plant emergence. The sowing density in the mixtures corresponded to half of the sole crops for both species, which corresponds to a substitutive design (Jolliffe, 2000). Seeds of both Purvalanol A were mixed within the row to promote species interactions. Experiment 2 was performed in a complete three randomized block design. Since it was impossible to evaluate all combinations of the 34 species in a bispecific legume/non-legume mixture, we selected five legumes such as: crimson clover (Trifolium incarnatum L.), purple vetch (Vicia benghalensis L.), wild lentil (Lens nigricans (M.Bieb.) Godr.), forage pea (Pisum sativum L.) and faba bean (Vicia faba L.); and five non-legumes, such as: turnip rape (Brassica rapa var. oleifera DC.), foxtail millet (Setaria italica (L.) P. Beauv.), bristle oat (Avena strigosa Schreb.), Italian ryegrass (Lolium multiflorum Lam.) and phacelia (Phacelia tanacetifolia Benth.), according to their known different characteristics, i.e. roots and shoot architectures, phenology, height and growth rate, for producing a broad range of interspecific interactions. We intercropped each legume with each non-legume to obtain 25 different mixtures. This experimental design of 25 mixtures was replicated in Experiment 3 at another experimental site, very different in terms of soil, climate and soil N availability, and performed in a two block randomized design instead of three blocks initially planned because of technical limitations. Experiment 4 was performed in 2013 in another experimental field with three randomized blocks, where 16 promising mixtures were sown. These mixtures were identified from the 2012 experiments has potentially well composed and were based on 14 species of legumes and non-legumes such as: crimson clover, Egyptian clover (Trifolium alexandrinum L.), purple vetch, common vetch (Vicia sativa L.), faba bean, and forage pea, turnip rape, white mustard (Sinapis alba L.), Abyssinian mustard (Brassica carinata A. Braun.), foxtail millet, forage sorghum (Sorghum bicolor Moench var. sudanense [Piper] Stapf.), bristle oat, Italian ryegrass, and phacelia. Data from Experiments 2–4 were used to calibrate the legume and non-legume sub-models of the mixture models. They were also used to validate the complete model built by summing the two “independent” sub-models and the legume proportion in the mixture.